The Debate That Arrived Too Late: Same Field, Different Realities in Mental Health AI

Mental health professionals are debating whether artificial intelligence should be permitted to operate in mental health contexts. This debate is the wrong question. Platforms such as Character.AI, Replika, and Wysa collectively reach tens of millions of users monthly for mental health conversations with no clinical governance framework applied. General-purpose AI assistants, not designed or marketed as mental health tools, handle mental health conversations at a scale that, by all available usage indicators, dwarfs every dedicated mental health platform combined. This has not happened through coordinated public health planning. It has happened through commercial deployment, user demand, and the emergent reality that always-available conversational AI is what people reach for when distress arrives at three in the morning and nothing else is accessible. Alongside this commercial reality, governments across major jurisdictions have made deliberate infrastructure decisions: Singapore has announced a national AI health strategy, the UK government has stated its ambition to make the NHS the most AI-enabled health system in the world, and the EU AI Act creates binding legal obligations for AI operating in health domains. The governance question is not whether AI should exist in this space but what it is doing, on what foundations, with what accountability, and for whom. This paper identifies two compounding reasons why current debates keep failing to address that question. First, existing regulatory frameworks were designed for objects with fixed identities (pharmaceutical drugs, medical devices, credentialled professionals) and cannot govern systems that are trained rather than programmed, simultaneously present everywhere, continuously changing, and structurally incapable of bearing accountability within existing legal frameworks. Second, and less recognised, participants in these debates are arguing from different locations on the mental health spectrum without acknowledging it. Clinicians working at the high-acuity end and public health practitioners working at the population wellbeing end are both accurately describing their portion of reality while treating it as the whole field. This produces debates that appear to be about values or evidence but are actually about different realities described through the same vocabulary. A three-domain governance framework is proposed that maps onto the full spectrum of human health and wellbeing, from population flourishing through clinical practice to the structural conditions that determine population mental health outcomes. The framework does not resolve debates by choosing between clinical and public health priorities. It reveals that those priorities apply to different domains, require different governance actors, and demand different accountability structures. Understanding this architecture is the prerequisite for governance conversations that can actually advance.